Unclean disconnect of stopping slave HC from the domain

An interesting problem popped up when executing a new test I wrote.[1]

The test invoked DomainLifecycleUtil.stop() to stop the slave HC. After it returned, it invoked an op on the master, one that would result in domain wide rollout. That failed with a failure indicating a problem sending the op to the slave.

But the slave should have been disconnected from the master before DomainLifecycleUtil.stop() returned. It does a process.stop() on the PC, which should cleanly stop the PC, which cleanly stops the HC, which should cleanly disconnect from the master as part of stopping. But somehow that didn't all happen.

[1] These specific details will soon be out of date; i.e. I'll change the test to work around this and eventually the test job will come out of TeamCity history. But anyway here they are:

https://ci.wildfly.org/viewLog.html?buildId=123039&tab=buildResultsDiv&buildTypeId=WildFlyCore_MasterIgnorWindows

java.lang.AssertionError: {"host-failure-descriptions" => {"slave" => "Writes closed"}}
	at org.junit.Assert.fail(Assert.java:88)
	at org.jboss.as.test.integration.domain.management.util.DomainTestSupport.validateResponse(DomainTestSupport.java:229)
	at org.jboss.as.test.integration.domain.management.util.DomainTestSupport.validateResponse(DomainTestSupport.java:221)
	at org.jboss.as.test.integration.domain.suites.HostLifecycleWithRolloutPlanTestCase.addRolloutPlan(HostLifecycleWithRolloutPlanTestCase.java:189)
	at org.jboss.as.test.integration.domain.suites.HostLifecycleWithRolloutPlanTestCase.testSlaveBootWithMissingRolloutPlan(HostLifecycleWithRolloutPlanTestCase.java:167)